Anthropic's Claude Gains Autonomous Scientific Capabilities and Expanded Enterprise Agent Tools

Anthropic tested Claude’s ability to design protein binders from scratch, a task representative of the early parts of the drug design process that has historically taken a specialist weeks or months per target. [3] Without human input into any design decision, Claude Opus 4.8 and Mythos Preview ran 24- to 48-hour campaigns against 16 targets. [1] They researched each target, chose epitopes, installed and ran open-source protein design and structure-prediction models, optimized their candidates in silico, and delivered 30 ranked designs per target. [1] The models designed protein binders against 15 targets, and succeeded against 14 of them. [3] Depending on the setup, between 22% and 35% of its individual designs bound successfully, compared to the 10-15% that is typical in protein design campaigns today. [3] Two independent contract research organizations synthesized every design exactly as delivered and measured its binding. [1]

In analytical chemistry, Claude Opus 5 was given NMR and LC-MS data to assess the identity and purity of compounds. [3] Provided with only a contract lab’s raw files and a two-sentence prompt, Claude returned finished results in 23 and 19 minutes. [3] The model matched the lab’s own analysis on hydrogen counts and purity, scoring 96.4% versus 96.33%. [3]

Microsoft Foundry offers Claude models in a version that runs on Azure infrastructure end-to-end. [2] When Claude models became generally available in Microsoft Foundry in June 2026, the launch included Entra ID authentication and Azure Marketplace billing. [5] This allowed teams to run Claude inside the same subscription, network perimeter, and cost-management surface as the rest of their Azure estate. [5]

Five new capabilities are now available for Claude models in Microsoft Foundry hosted on Azure: Structured outputs, Web search, Web fetch, MCP connector, and Tool search. [5] These features serve as the building blocks that turn a model endpoint into a production agent platform. [5] The release moves bespoke engineering scaffolding, such as a hand-rolled MCP client or a tool router, directly into the platform on deployments hosted on Azure. [5]

The integration of Claude into Microsoft Foundry's secure Azure environment removes major procurement and governance obstacles for enterprise teams. Simultaneously, Anthropic's demonstration of autonomous protein design highlights a massive leap in applied scientific AI, pushing hit rates up to 35% compared to the industry standard 10-15% for de novo design campaigns. By pairing raw reasoning power in models like Opus 5 with native agent scaffolding—such as MCP connectors and structured outputs—Anthropic and Microsoft are transitioning LLMs from single-call endpoints into comprehensive, automated workflows. The ability to autonomously research targets, install open-source software, and evaluate candidates in a single session represents a significant shift from isolated copilots to autonomous agents. What the sources don't address: how much compute these 48-hour autonomous scientific campaigns cost to execute.